Partition Move

Currently Partition Move only implements the Kernighan-Lin-Algorithm. More...

Inheritance diagram for opengm::PartitionMove< GM, ACC >:

Collaboration diagram for opengm::PartitionMove< GM, ACC >:

Public Types
typedef ACC	AccumulationType

typedef GM	GraphicalModelType

typedef size_t	LPIndexType

typedef visitors::VerboseVisitor< PartitionMove< GM, ACC > >	VerboseVisitorType

typedef visitors::EmptyVisitor< PartitionMove< GM, ACC > >	EmptyVisitorType

typedef visitors::TimingVisitor< PartitionMove< GM, ACC > >	TimingVisitorType

typedef __gnu_cxx::hash_map< IndexType, LPIndexType >	EdgeMapType

typedef __gnu_cxx::hash_set< IndexType >	VariableSetType

Public Types inherited from opengm::Inference< GM, ACC >
typedef GM	GraphicalModelType

typedef ACC	AccumulationType

typedef GraphicalModelType::LabelType	LabelType

typedef GraphicalModelType::IndexType	IndexType

typedef GraphicalModelType::ValueType	ValueType

typedef GraphicalModelType::OperatorType	OperatorType

typedef GraphicalModelType::FactorType	FactorType

typedef GraphicalModelType::IndependentFactorType	IndependentFactorType

typedef GraphicalModelType::FunctionIdentifier	FunctionIdentifier

Public Member Functions
	~PartitionMove ()

	PartitionMove (const GraphicalModelType &, Parameter para=Parameter())

virtual std::string	name () const

const GraphicalModelType &	graphicalModel () const

virtual InferenceTermination	infer ()

template<class VisitorType >
InferenceTermination	infer (VisitorType &)

virtual InferenceTermination	arg (std::vector< LabelType > &, const size_t=1) const
	output a solution More...

Public Member Functions inherited from opengm::Inference< GM, ACC >
virtual	~Inference ()

virtual void	setStartingPoint (typename std::vector< LabelType >::const_iterator)
	set initial labeling More...

virtual InferenceTermination	args (std::vector< std::vector< LabelType > > &) const

virtual InferenceTermination	marginal (const size_t, IndependentFactorType &) const
	output a solution for a marginal for a specific variable More...

virtual InferenceTermination	factorMarginal (const size_t, IndependentFactorType &) const
	output a solution for a marginal for all variables connected to a factor More...

virtual ValueType	bound () const
	return a bound on the solution More...

virtual ValueType	value () const
	return the solution (value) More...

InferenceTermination	constrainedOptimum (std::vector< IndexType > &, std::vector< LabelType > &, std::vector< LabelType > &) const

InferenceTermination	modeFromMarginal (std::vector< LabelType > &) const

InferenceTermination	modeFromFactorMarginal (std::vector< LabelType > &) const

Detailed Description

Partition Move

Currently Partition Move only implements the Kernighan-Lin-Algorithm.

Cite: B.W. Kernighan and S. Lin, "An efficent heuristic procedure for partition graphs", 1970
Maximum factor order : second order Potts functions only!
Maximum number of labels : same as the number of variables !
Restrictions : see above
Convergent : Converge to some local fix point

Definition at line 39 of file partition-move.hxx.